Efficient and Effective Multi-person and Multi-angle Face Recognition based on Deep CNN Architecture

Abstract

Recently, the development and application of robots has become a famous topic and the most important thing for robots is personification. This paper uses a webcam to capture the image as a visual system input. The facial image is obtained through high performance face detect neural network. Facial landmarks are used to correct the face. Then, we use facial color RGB images for facial feature detection and identity recognition. By training a complete feature detection network, it is possible to detect valid and distinct facial features and train the classifier for those features. We can obtain identity confidence by using classifier for those feature. The experiment results show that the accuracy of identity recognition can be as high as 90.61%. In practical applications, the system can recognize identities up to thousands of people at the same time.